Claude Opus 4.8 vs GPT-5.2
Wins 6 of 8 areas
Coding · Agents · Reasoning · Facts · Images and charts · Math
Wins 1 of 8 areas
Long documents
Claude Opus 4.8 is the stronger all-rounder.GPT-5.2 is cheaper and better at long documents.
Scores updated · 52 tests both models report · How we compare
Where each one wins
Tests won in each of the eight areas we test. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
- CodingWriting and fixing software70Claude Opus 4.87 of 8 tests · 1 tie
- ReasoningHard problems that need careful thinking60Claude Opus 4.86 of 6 tests
- AgentsCarrying out multi-step tasks on its own40Claude Opus 4.84 of 5 tests · 1 tie
- FactsGetting facts right instead of making them up30Claude Opus 4.83 of 3 tests
- MathCompetition and research-level math42Claude Opus 4.84 of 6 tests
- Images and chartsUnderstanding pictures, charts and video20Claude Opus 4.82 of 2 tests
- Long documentsFinding answers in very long texts01GPT-5.21 of 1 test
- Following instructionsDoing exactly what it is asked11Even1 each
Long documents rests on a single test.
What it costs
Prices per million tokens, roughly 750,000 words. The bars show the cost of a million tokens read plus a million written.
GPT-5.2 costs 48% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where Claude Opus 4.8 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+41.9points ahead
- Long, multi-file coding tasks in real codebasesSWE-bench Pro+39.3points ahead
- Extremely hard research-level math problemsFrontierMath Tier 4+24.4points ahead
Where GPT-5.2 pulls ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+13.2points ahead
- Reasons across sets of long documentsAA-LCR+5points ahead
- Olympiad math problems with short answersIMOAnswerBench+2.8points ahead
Every test, side by side
All 52 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingClaude Opus 4.8
- SWE-bench ProClaude Opus 4.8 by 39.369.229.9+39.3
- SWE-bench MultilingualClaude Opus 4.8 by 17.784.466.7+17.7
- LMArena · WebDevClaude Opus 4.8 by 141 rating points15561415+141 rating
- Terminal-Bench HardClaude Opus 4.8 by 11.358.347+11.3
- SWE-bench VerifiedClaude Opus 4.8 by 8.688.680+8.6
- LiveBench · CodingClaude Opus 4.8 by 5.781.876.1+5.7
- SciCodeClaude Opus 4.8 by 2.354.452.1+2.3
- LiveBench · Agentic Codingtie50.550.3tie
AgentsClaude Opus 4.8
- τ-Bench V3 · BankingClaude Opus 4.8 by 23.134.211.1+23.1
- MCP AtlasClaude Opus 4.8 by 21.682.260.6+21.6
- BrowseCompClaude Opus 4.8 by 18.584.365.8+18.5
- AA ApexAgentsClaude Opus 4.8 by 16.439.423+16.4
- GDPValtie47.848.3tie
ReasoningClaude Opus 4.8
- ARC-AGI-2Claude Opus 4.8 by 19.272.152.9+19.2
- SimpleBenchClaude Opus 4.8 by 1964.845.8+19
- Humanity's Last ExamClaude Opus 4.8 by 1148.737.7+11
- CritPtClaude Opus 4.8 by 9.320.911.6+9.3
- LiveBench · ReasoningClaude Opus 4.8 by 689.283.2+6
- GPQA DiamondClaude Opus 4.8 by 1.79290.3+1.7
FactsClaude Opus 4.8
- AA-Omniscience · Non-hallucinationClaude Opus 4.8 by 41.960.718.8+41.9
- SimpleQA VerifiedClaude Opus 4.8 by 15.95337.1+15.9
- AA-Omniscience · AccuracyClaude Opus 4.8 by 4.548.844.3+4.5
Images and chartsClaude Opus 4.8
- CharXiv (RQ)Claude Opus 4.8 by 7.889.982.1+7.8
- LMArena · VisionClaude Opus 4.8 by 49 rating points12951246+49 rating
MathClaude Opus 4.8
- FrontierMath Tier 4Claude Opus 4.8 by 24.456.131.7+24.4
- FrontierMath Tiers 1-3 (v2)Claude Opus 4.8 by 12.68067.4+12.6
- IMOAnswerBenchGPT-5.2 by 2.883.586.3+2.8
- AIME 2026Claude Opus 4.8 by 1.710098.3+1.7
- HMMT Feb. 2026GPT-5.2 by 1.595.597+1.5
- LiveBench · MathematicsClaude Opus 4.8 by 1.194.393.2+1.1
Following instructionsEven
- IFBenchGPT-5.2 by 13.262.275.4+13.2
- LiveBench · Instruction FollowingClaude Opus 4.8 by 10.27261.8+10.2
Other results19 tests, not counted
Tests outside the eight areas. They are not counted above: several are summary scores built from other tests, or the same test under another name.
- GDPval-AA (Elo)Claude Opus 4.8 by 428 rating points18901462+428 rating
- AA-OmniscienceClaude Opus 4.8 by 29.728.8-0.9+29.7
- Artificial Analysis Coding IndexClaude Opus 4.8 by 25.674.348.7+25.6
- DeepSearchQA (F1)Claude Opus 4.8 by 21.893.171.3+21.8
- Terminal-Bench 2.0Claude Opus 4.8 by 20.674.654+20.6
- AA Agentic IndexGPT-5.2 by 17.642.660.2+17.6
- PaperBenchClaude Opus 4.8 by 16.680.363.7+16.6
- ToolathlonClaude Opus 4.8 by 13.659.946.3+13.6
- frontiermath_tier_4_v1Claude Opus 4.8 by 12.531.318.8+12.5
- HLE (with tools)Claude Opus 4.8 by 12.457.945.5+12.4
- livebench_data_analysisGPT-5.2 by 12.26678.2+12.2
- AA IntelligenceClaude Opus 4.8 by 11.441.830.4+11.4
- τ²-Bench Telecom (AA run)Claude Opus 4.8 by 9.694.484.8+9.6
- ARC-AGI-1Claude Opus 4.8 by 6.392.586.2+6.3
- matharena_visual_math_overallGPT-5.2 by 4.981.686.5+4.9
- LiveBenchClaude Opus 4.8 by 2.477.274.8+2.4
- ScreenSpot-Pro (No tools)Claude Opus 4.8 by 1.687.986.3+1.6
- HMMT Nov. 2025tie96.595.8tie
- livebench_languagetie79.779.8tie
Questions people ask
Which is better, Claude Opus 4.8 or GPT-5.2?
Claude Opus 4.8 wins six of the eight areas we test: coding, agents, reasoning, facts, images and charts and math. GPT-5.2 wins long documents, and costs 48% less. They are level on following instructions.
Which is better for coding?
Claude Opus 4.8. It wins 7 of the 8 coding tests both models report; GPT-5.2 wins none, and 1 is a tie.
Which is cheaper?
Claude Opus 4.8 costs $5.00 per million input tokens and $25.00 per million output tokens; GPT-5.2 costs $1.75 and $14.00. That makes GPT-5.2 about 48% cheaper for the same work.
How do you compare the two?
We use the 52 benchmark tests both models have published scores on. The verdict counts the 33 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 19 are listed but not counted, because several are summary scores or repeat a test. Each score is the one shown on the model's own page.